Understanding Voters' Expectations and Behaviour: A Best-Worst Analysis. Abstract. Introduction

Transcription

1 Understanding Voters' Expectations and Behaviour: A Best-Worst Analysis Dr. Herve Remaud, Michael Gillan, University ofsouth Australia Abstract Floating and undecided voters are decisive voters for any political party as late decision making could change the outcome of the election. Therefore, better understanding these individuals' expectations is critical in order to reach them at the right time with the right words. In this study, we measured the degree of importance of 16 policies proposed by two major political parties for the Assemblee Nationale election in France (held in June 2007) using Best Worst Scaling approach. Voters who expected to vote for the UMP party were more in line with this party's program compared to voters for its rival party (the PS). Undecided voters were especialiy responsive to policies on poliution. Our findings suggest that the PS would have gained more votes with a better coverage ofthat topic during the days prior to the election and by promoting the PS and its program more effectively. Introduction Winning an election is an important objective for ali politicians. Some authors recently suggested that being more market-orientated may help to understand voters' expectations and thus provide a means of winning the votes of a smaii number of swinging voters in marginal seats (Baines et al., 2005; Worcester and Mortimor, 2005). While some electors are faithful to a specific ideology and would never change their vote for a party, others are floating voters and are likely to decide for whom they will vote the days or the few hours prior their vote. The question raised by this situation is: do undecided voters have different political expectations compared to voters who make up their mind earlier in the election campaigning? Respond to this research question highlights two methodological issues. The first one is linked to the expectations or political policies that should be assessed. In other words, which ones should be included in the study? The second one is more methodological: how should these expectations be assessed? Most of the studies undertaking in marketing use a Likert scale approach in order to assess some statements. The method is very easy to use and quite cost effective. However, when asking individuals to select their preferences among a list of statements, the Likert scale approach does not force people to compare these statements. Another way to assess relative preferences of individuals is to use the Best Worst Scaling (named BWS) approach developed by Louviere and his colleagues (Finn and Louviere, 1992; Marley and Louviere, 2005). We use that approach to measure the importance of selected political policies for the election ofthe Assemblee Nationale (held in France in June 2007). The paper is structured as follows. We start with a brief literature review about political marketing, highlighting the importance of understanding undecided voters' behaviour and expectations. We then present the research approach used to assess political expectations and the main findings. We finally discuss the implications of using BWS approach for politicians and in a political marketing perspective. 1160

2 Theoretical Background Political Marketing is "the science ofhow politicians interface with their electorates" (Lilleker and Lees-Marshment, 2005); it is concerned with communicating with party members, media and prospective sources of funding as well as the electorate (Lock and Harris, 1996). One big issue facing western democracies is low voter turnout (Dermody and Hanmer-Lloyd, 2004), and declines in party membership (Johansen, 2005), which both create difficulties for political parties to size electors up and understand their critical expectations. Generally the electorate is uninterested in politics, which makes individual voters difficult to reach during a political campaign (Lock and Harris, 1996). Therefore, several voting behaviour theories have been developed, though none fully explain voting behaviour (Bannon, 2003). In the meantime, as forecasting an election outcome is not completely accurate (Lewis-Beck, 2005), politicians have to find other ways to maximise their chance of being elected. One of the major issues and concerns for political parties and politicians is to assess what undecided and floating electors expect from politicians and political parties. This is crucial in an election as these electors' support may be critical to success as suggested by Belenky and King (2007). During the last weeks and days before the actual election, these electors would be under the influence ofother electors but would also be open to influence from marketing campaigns and politicians' speeches. Having the ability to deliver the right message at the right time at the right place may change the outcome ofan election. The question raised by this situation is how to assess the political expectations of electors in general and undecided voters in particular. In other words, do undecided voters have different political expectations when compared to voters who already know the party for which they will vote? In essence, the answer would be yes when voters are confronting a list of policies belonging to one or the other major political party in contention for an election. But undecided voters may prefer some policies from one party yet others from another party. In brief, is there a voting behavior theory that works to explain undecided voters' behavior? Following the presidential election in France, the Assemblee Nationale election was viewed as one way in which the new President could exercise his mandate. In that sense, UMP is fully promoted during this election, more as a party than as one single or specific candidate. One can suggest that party identification theory could be used to explain political (un)success. However, party identification theory may be modified by the leader's profile effect in different local areas. Some voters may identify with one specific party but will avoid voting for someone they cannot support. Party identification should also be modified by issue voting. In order to not give all power to the UMP party (the party ofthe new President), other parties may have encouraged people to not vote for UMP candidates; this has been one way for the PS to encourage voters to vote for one PS candidate. However, some voters may support some policies proposed by one party and others from another party. On the other hand, one can assume that political parties have an (good) understanding of their members' and supporters' expectations. Focus groups and internal surveys can easily be conducted to understand and highlight these expectations. Based on that assumption, the following hypotheses are proposed: Hla: Voters who declare that they expect to vote for the Parti Socialiste (PS) are more likely to rate political measures proposed by this party highly than voters who plan to support a different party. 1161

3 Hlb: Voters who declare that they expect to vote for the Union pour la Majorite Presidentielle (UMP) are more likely to rate political measures proposed by this party highly than voters who plan to support a different party. Because less consistency is expected among undecided voters, the following hypothesis is proposed. H2: Voters who declare that they are still not decided to vote for a specific political party are likely to exhibit higher levels of variation in their support for political measures when compared to voters who already know the party for which they will vote. We used the Best Worst Scaling approach to measure the importance of selected political policies for the election of the Assembtee Nationale (held in France in June 2007). BWS has been developed by Louviere and his colleagues (Finn and Louviere, 1992; Marley and Louviere, 2005). BWS model assumes that respondents behave as ifthey are examining every possible pair in each subset and then they choose the most distinct pair as the best-worst, most-least, maximum difference pair (Cohen and Neira, 2003). That means that the BWS model requires respondents to make trade offs among benefits. BWS is a more discriminating way to measure attribute importance than either rating scales or the method of paired comparison (Cohen, 2003). BWS approach has been applied to a various range of topics: Auger et al. (2007) investigated consumer beliefs towards social and ethical issues across six countries; Flynn et al. (2006) applied BWS to answer question such as "Is quality of care more highly valued that waiting time?"; Goodman et al. (2005) analysed wine consumer choice using this approach. This study represents the first application of BWS to political marketing. Research Approach The political policies used in this exploratory study have been promoted by the major parties engaged in the presidential / Assembtee Nationale election race: S. Royal for the Parti Socialiste, and N. Sarkozy for the 'Union pour la Majorite Presidentielle. The following selected policies 'belong to' the Parti Socialiste: Invest in renewable energy sources to attain 20% of energy consumption by 2020; Apply the principle of 'polluter - pays' to firms responsible for pollution; Permanently increase financing of public hospitals; Fight against school-yard bulling by increasing adult presence; Introduce policies to avoid households overborrowing; One single political mandate at the time; Increase the guaranteed minimum wage to 15000; Implement programs to reduce obesity; The following ones 'belong to' the Union pour la Majorite Presidentielle: Reduce income tax; Allow tax concessions on overtime payments; Renew the nuclear energy program; Allow free parking in cities to clean motor vehicles; Reinforce selected immigration policy; Introduce high speed Internet to all French; Evaluate and publish university results. The sample used was a convenience sample. The invitation to participate was sent to more than 300 people with various backgrounds. Half of this population were staff members (academic and administrative staff) based at three different University research centres in economic and management located in the South of France. The other half of the sample was 1162

4 personal contacts ofthe researchers. The total size ofthe population is about 320 people. As 102 people completed the survey 1, the response rate was about 31.8%. The survey was conducted using a web-based questionnaire during the 10 days prior the first round of the election, i.e. from the first of June to the 10 of June Other questions included demographic variables, voters' behaviour for the last presidential election, and voters' expected behaviour for the Assemblee Nationale election. Results Figure 1 presents voters' expectations regarding the 16 political policies selected for this study. Best-Worst scores have been standardised into a scale from n to 100, with each representing the probability to choose one political policy as most important, and where all other standardised scores are relative to the most important chosen policy. Most versus Least Important political policies expected to be Implemented by the new Assembhle Natlonale -' -:;; ~:;~i~ I"!'~!""!'~C!'~"!""!'''!'"'!=!''=!''!'=!''!~_!'=!'"=!"~~~~~54~'9~[FI:6:9;'6::=I::~90~'4~100 I htroduce polcies to avoil househok:!s over-borrow ilg I 51.7 Ole si1gle poweal rtbndate atthe tirre ncreas8 the guaranteed ninimjmwage (to 1500 ) 369 I Renew the nuclear energy program 35 ; RIlb1forc.S.Jecl~m::::=::."...,,.,;... ' c'... "'26 2 ;'1,1. "E!' --'E!E!...;a.. _.: ': 23.6 I' AIbw tax concessions on overtlrre payrrents ",,=E!E!"" AlkJw free par1dng In dies to clean rmtor vehicles _~ 17.6 ltl>lef11!nl progr8tr6 10 reduc8 obesity = ,.... ntroduce high speed ~I.rnello 81 French 13.6 Evaluate and publsh universly results Give tax cleductions to union subscriptions 0.0 r o ~-~---=~--~-~-~-~--~-~-~ Figure 1: Voters' Expectations for the Assemblee Nationale Election When analysing all respondents' responses, "Invest in renewable energy sources to attain 20% ofthe energy consumption by 2020" and "Permanently increase financing ofpublic hospitals" were the two expectations the voters were more likely (score of 100 and 90) to choose as most important among this list (there is no statistical difference between the two scores using LSD test). In the middle of the scale, "Introduce policies to avoid households over-borrowing" is half as likely (score of 51.7) to be chosen as the most important political policy to be implemented by the new Assemblee Nationale. Table 1 highlights voters' expectations linked to their expected voting behaviour 2 for this election (i.e., scores were significantly different using LSD statistical test). When splitting the I A briefoverview ofthis sample shows that: 55% ofthe respondents are female and 45% are male; 46% are between 18 to 34 years old, 26.5% are years old, and 27.5% are more than 44 years old; and regarding the net monthly income, 27.5% earn less than 1,000 euros, 34.3% earn between 1,000 and 2,000 euros, and 38.2% earn more than 2,000 euros. 2 For the next election, I would vote for a candidate ofthe PS; UMP; Other Parties; Blank; I will not vote; or Still not decided 1163

5 sample between PS, UMP, Other Parties and Undecided voters (Table 1), voters showed differing levels ofsupport for 6 policies among the 16 proposed. Table 1: Voters' Expectations and Behaviour for the Assemblee Nationale Election Reinforce selected immigration policy (UMP politicalprogram) Other Renew the nuclear energy program (UMP politicalprogram) Apply principle of' poliuter - pays I to firms responsible for poliution (PS politicalprogram) Allow tax concessions on overtime payments (UMP political program) Reduce income tax (UMP political program) Permanently increase financing ofpublic hospitals (PS political program) 32.9 (no sig. dif) 35.9 (no sig. dif) 82.7 (no sig. dif) Discussion and Conclusion Voters who declared they would vote for an UMP candidate were more likely to rate four UMP party policies highly (compared to other voters and out ofthe 7 UMP policies included in the study). Those voters were less likely to rate a rival PS party pol icy highly (Permanently increase financing of public hospitals). This provides partial support for hypothesis HIband suggests UMP supporters identify strongly with the UMP party, and are more likely to vote for the UMP party in the polling booth. On the other hand, the findings shown in Table 1 do not support hypothesis Hla. Voters who expect to vote for a PS candidate did not rate policies proposed by this party more highly. Applying the principle of 'polluter - pays' to firms responsible for pollution was almost more highly rated by UMP expected voters. The other PS policy was seen as important by voters of other parties and undecided voters. These findings suggest a weak identification ofps voters with the PS party or some inefficiencies for this party to promote some policies of its program. Undecided voters were more likely to choose the policy "Apply principle of 'polluter - pays' to firms responsible for pollution" as most important (score of92.7) to be implemented by the new Assemblee Nationale compared to scores ofaround 55 for all other voters. For two UMP policies, undecided voters were more likely to differ from decided voters, though for the last 3 policies included in Table 1, undecided voters did not have significantly different ratings compared to other voters. Overall, H2 was marginally supported by the data. Based on these findings, UMP supporters have a clear understanding of what the party is currently promoting. On the other hand, PS expected voters seem to not have a clear understanding of what the party is promoting, and undecided voters do not behave differently compared to other voters, except for one specific policy in the study. More than promoting (even more) this policy, one can suggest that a clearer and more coherent positioning of the PS would have given this party more votes from undecided voters. This research project has some limitations: the size and structure of the sample, the choice of policies to be assessed, the period of time used to interview voters, and the extent to which 1164

6 leadership or candidate profiles would affect voters' behaviour. Most of these limitations are intrinsic to exploratory studies. However, this study demonstrates that the BWS approach can be successfully used to better understand voters' political expectations and it thus opens a new stream for future research projects in political marketing. 1165

7 References Auger, P., Devinney, M.T., Louviere, J.1., Using best-worst scaling methodology to investigate consumer ethical beliefs across countries. Journal ofbusiness Ethics 70, Baines, P., Worcester, R., Jarret, D., Mortimore, R., Product attribute-based voter segmentation and resource advantage theory. Journal ofmarketing Management 21 (9-10), Bannon, P., Voting, non-voting and consumer buying behaviour: Non-voter segmentation and the underlining causes ofelectoral inactivity. Journal ofpublic Affairs 3 (2), Belenky, A.S., King D.C., A mathematical model for estimating the potential margin of state undecided voters for a candidate in a US Federal election. Mathematical and Computer Modelling 45, Cohen, S.H., Maximum difference scaling: Improved measures ofimportance and preference for segmentation. Sawtooth Software, Research Paper Series. Cohen, S.H., Neira L., Measuring preference for product benefits across countries. overcoming scale usage bias with maximum difference scaling. ESOMAR Conference, Latin American Conference, Uruguay. Dermody, J., Hanmer-Lloyd, S., Segmenting youth voting behaviour through trustingdistrusting relationships: A conceptual approach. International Journal ofnonprofit & Voluntary Sector Marketing 9 (3), Flynn, T.N., Louviere, J.J., Determining the appropriate response to evidence ofpublic concern: The case offood safety. Journal ofpublic Policy & Marketing 11 (1), Flynn, T.N., Louviere, J.1., Peters, T.1., Coast, J., Best-worst scaling: What it can do for health care research and how to do it? Journal ofhealth Economics 26 (1), Goodman, S., Lockshin, L., Cohen, E., Best-worst scaling: a simple method to determine drinks and wine style preference. Proceedings 2nd International Wine Marketing Symposium, Sonoma CA. Johansen, H., Political marketing. Journal ofpolitical Marketing 4 (4), Lewis-Beck, M.S., Election forecasting: Principles and practice. British Journal of Politics & International Relations 7, Lilleker, G., Lees-Marshment, 1., Political marketing: a comparative perspective, Manchester University Press, Manchester, New York. Lock, A., Harris, P., Political r;narketing - vive la difference! European Journal of Marketing 30 (10-11),

8 Marley, A.AJ., Louviere, J.1., Some probabilistic models ofbest, worst, and best-worst choices. Journal ofmathematical Psychology 49, Newman, RI., Editorial: Broadening the boundaries ofmarketing: Political marketing in the new millennium. Psychology and Marketing 19 (12), Worcester, R.M., Mortimor, R., Political triangulation: Measuring and reporting the key aspects ofparty and leader standing before and during elections. Journal ofpolitical Marketing 4 (2-3),